I am trying to "normalize" accelerometer data for a sensor that is not centered on a car. In the image below, you can see how centripetal force changes based on location.

In my case, I only have one sensor on the right side of the vehicle. For logistical reasons, I can only have one sensor and it must be on the right hand side. As I turn, the force is more exaggerated turning left compared to turning right.

Assuming I know the exact dimensions of the car and the exact location of the sensor, is there a way to normalize the data as if the accelerometer were centralized (in other words, left and right turns would have identical signatures)? Assume there is no noise, that we are pre-smoothing the signal.

enter image description here


With only the readings of one accelerometer you're out of luck.

The basic reason is that a particular reading could be due to a high-speed, large-radius turn or a low-speed, small-radius turn and the two cases call for different corrections.

If you have secondary data—such a ground velocity (speedometer speed is only approximately correct for this but you could make do) or angular velocity (from a compass or gryoscope)—then you can work something out.

Of course, for smooth driving behaviors you could also get enough secondary data from GPS, but in that case you don't actually need the accelerometer at all.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.